TL;DR
We’re outsourcing more of our thinking to AI and digital tools than ever before. Research shows this changes how we encode memories, potentially weakens critical thinking, and may even shrink parts of our brain. But before you throw your phone in a lake, the story is more nuanced. This isn’t just another tech panic. It’s a look at what actually happens when we let machines do the cognitive heavy lifting, backed by neuroscience, psychology, and some surprisingly hopeful counterarguments.
Introduction
Here’s something that should make you uncomfortable: when people expect Google to remember something for them, their recall drops by nearly 30%. That’s not a guess. That’s from a 2011 study at Columbia University that kicked off a decade of research into what happens when we stop using our brains [1].
And it gets weirder. GPS users don’t just navigate worse without it. They actually have smaller hippocampi [2]. The part of your brain responsible for spatial memory physically shrinks when you let your phone tell you where to turn.
Now we’re handing over not just navigation and trivia, but reasoning itself to AI. ChatGPT writes our emails. Claude debugs our code. DeepSeek does our math homework. And a growing body of research suggests we might be trading convenience for something we can’t get back: the ability to think deeply.
But maybe that’s fine? Maybe cognitive offloading is just the next step in human evolution. After all, Socrates thought writing would ruin our memories [3], and here we are, thousands of years later, doing pretty okay.
So which is it? Are we dumbing ourselves down, or are we just evolving how we think?
When Google Became Your External Hard Drive
Betsy Sparrow didn’t set out to prove we’re all getting dumber. She just wanted to know what happens when people think the internet will remember things for them.
Turns out, plenty happens.
In her 2011 study published in Science, participants read trivia statements and were told the computer would either save them or delete them [1]. When people thought the facts would be saved, they barely remembered them. When they thought the facts would be erased, recall shot up. The difference wasn’t small. It was the kind of gap that makes you wonder if your brain is just being lazy.
But here’s the kicker: when information was saved to specific folders, people got really good at remembering where it was stored. Not what it was. Just where to find it.
Sparrow called this the „Google effect“ [4]. Your brain treats the internet like a coworker who’s really good at trivia. You don’t bother memorizing what they know. You just remember that they know it.
This builds on Daniel Wegner’s concept of „transactive memory systems“ [5]. In relationships and teams, people naturally divide up who remembers what. You don’t need to know your partner’s mom’s birthday if they always remember it. You just need to know that they know.
The internet has become humanity’s primary transactive memory partner. And unlike your spouse, it never forgets an anniversary.
A comprehensive 2024 meta-analysis examining 22 articles with 30,889 participants confirmed the effect is real and consistent [6]. It’s stronger on phones than computers. It hits North Americans harder than other regions. And people with more existing knowledge resist it better, like having a strong immune system protects you from getting sick.
Adrian Ward at UT Austin added another layer. He found that just having your phone nearby, even when it’s off and face-down, reduces your working memory and fluid intelligence [7]. The mental effort of not thinking about your phone drains cognitive resources. The people most dependent on their phones suffered the worst.
So yeah, Google didn’t just change how we find information. It rewired how we store memories.
Your Brain Isn’t a Muscle, But It Acts Like One
If you stop going to the gym, your muscles atrophy. Turns out, your brain works the same way.
Michael Merzenich won the Kavli Prize for mapping how the brain reorganizes itself based on experience [8]. His research showed that sensory cortex maps aren’t fixed. They constantly remodel based on what you do. Practice piano, and the finger regions expand. Stop practicing, and they shrink.
Here’s the part that should worry you: „It is just as easy to degrade the brain’s processing abilities as it is to strengthen or refine them“ [8].
Neuroplasticity goes both ways.
The mechanism is brutally efficient. Your brain tags low-activity synapses for removal. Microglia cells show up, mark them with proteins, and eat them. It’s called synaptic pruning, and it’s not metaphorical. Your brain literally deletes unused connections [9].
Eleanor Maguire’s famous study of London taxi drivers proves the point [10]. Cabbies who memorized 25,000 streets had significantly larger posterior hippocampi than the general population. The size correlated with years of experience. And when Maguire tracked trainees over four years, only the ones who passed „The Knowledge“ showed brain growth [11].
Every single one who qualified had measurable structural changes. Every one.
Now flip that. What happens when you stop navigating and let GPS do it for you?
Dahmani and Bohbot found that lifetime GPS use correlated with worse spatial memory during unaided navigation [2]. And here’s the critical part: people who used GPS more didn’t do it because they were bad at directions. GPS use predicted declining spatial memory over time.
The tool caused the decline. Not the other way around.
Then there’s neurogenesis. Your brain generates about 10,000 new neurons daily in the hippocampus. But more than half die within weeks unless you engage in effortful learning [12]. Only difficult tasks that require sustained concentration rescue those neurons from death.
So while machines degrade through use, biological systems strengthen through challenge. When we outsource cognitive work to technology, the neural circuits for those tasks may degrade, new neurons may not be rescued, and cognitive reserve may not accumulate.
Use it or lose it isn’t a motivational slogan. It’s neuroscience.
Why Struggle Is the Whole Point of Learning
Robert Bjork spent decades studying how people learn. His big insight? We’re terrible at knowing what’s good for us [13].
Conditions that make learning feel easy usually fail to support long-term retention. Conditions that create difficulty and slow progress often optimize retention and transfer.
He calls them „desirable difficulties“ [14].
In one study, kids practiced throwing beanbags at a target. Group A practiced only at the test distance. Group B practiced at variable distances, but never the test distance. Group B crushed Group A. They’d never even practiced the exact distance they were tested on.
Math students who interleaved different types of problems scored 63% correct on delayed tests. Students who practiced one type at a time scored 20%. Yet when asked, the blocked-practice group thought they’d learned better [13].
We mistake fluency for learning.
The generation effect makes this even clearer. Actively generating information produces better memory than passively reading it. The effect size is huge, confirmed by hundreds of studies [15]. The testing effect shows the same thing: retrieval practice beats restudying every time.
As Bjork puts it: „Any time you look up an answer or have somebody tell you something you could generate instead, you rob yourself of a powerful learning opportunity“ [14].
When AI gives you instant answers, it bypasses all of this. No semantic processing. No generation. No retrieval practice. Just smooth, easy, instantly forgotten information.
A University of Pennsylvania study nailed this paradox [16]. Turkish high school students used ChatGPT for math practice. They answered 48% more problems correctly during practice. Then they took a test on the underlying concepts.
They scored 17% lower than students who struggled without AI.
The AI made them better at doing problems. It made them worse at understanding math.
That’s not a bug. That’s what happens when you remove the struggle that drives learning.
TikTok Broke Your Attention Span (Probably)
A 2025 meta-analysis looked at nearly 100,000 people across 71 studies on short-form video [17]. The findings weren’t subtle.
Moderate negative correlation between short-form video use and cognitive performance. Attention showed the strongest hit (r = -.38), followed by inhibitory control (r = -.41). The pattern held across age groups and platforms: TikTok, Instagram Reels, YouTube Shorts.
The mechanism is dopamine. Social media exploits the mesolimbic pathway, the same circuit involved in addiction [18]. Stanford research found the biggest dopamine spike doesn’t come from getting a like. It comes from the uncertainty of whether you’ll get one.
Intermittent reinforcement is more powerful than consistent rewards. Slot machines figured this out decades ago. Now your phone has.
The 2009 Stanford study on media multitasking found that heavy multitaskers were worse at task-switching despite having more practice [19]. They were more distracted by irrelevant stimuli. Brain scans showed less gray matter in the anterior cingulate cortex, a region critical for cognitive control.
Cal Newport pulls it all together with his „deep work“ framework [20]. McKinsey found the average knowledge worker spends over 60% of the week on electronic communication and internet searching. Thirty percent on email alone [21].
Sophie Leroy’s research on attention residue shows that switching tasks leaves mental residue on the previous task [22]. It takes over 23 minutes to fully refocus after an interruption.
But here’s the catch: most of this research is correlational. We can’t prove TikTok causes attention problems versus attracting people who already had them. Few studies track changes over time. Bidirectional relationships remain possible.
As Dr. Poppy Watson notes, „The link between online content consumption and decreased cognitive ability is so far a correlative one“ [17].
Still. When you can’t watch a two-minute video without checking your phone, something is clearly wrong.
Your Brain Is a Lazy Bastard (And Technology Knows It)
Daniel Kahneman’s „Thinking, Fast and Slow“ explains why we’re so vulnerable to cognitive shortcuts [23].
System 1 operates automatically with low effort. System 2 requires deliberate engagement. And here’s the problem: „System 2 is a lazy controller and doesn’t like to expend much effort“ [23].
We substitute easy questions for hard ones. We accept System 1’s answers without scrutiny. We default to intuition.
Fiske and Taylor called humans „cognitive misers“ [24]. We solve problems using minimal cognitive resources. When technology removes cognitive demands, it prevents System 2 from engaging at all.
The research on calculators makes this concrete. Students who used calculators more showed worse fundamental math skills. One study found students performed 42% correct without calculators versus 82% with them [25]. Many couldn’t proceed without the tool at all.
That’s not augmentation. That’s dependency.
Spell-checkers show the same pattern. Students don’t internalize corrections made by spell-checkers. The effect is limited on the cognitive level [26]. Effort searching for correct spellings relates to better learning.
A 2025 Microsoft Research study found that knowledge workers with higher confidence in generative AI tend to apply less critical thinking to AI outputs [27].
And here’s the automation irony: by mechanizing routine tasks and leaving exceptions to humans, you deprive humans of opportunities to practice judgment. When the exception finally arrives, they’re unprepared [28].
We’re building systems that make us worse at the very skills we’ll need when those systems fail.
Maybe AI Is Just Another Pencil
Andy Clark and David Chalmers proposed the „extended mind“ thesis in 1998 [29]. Their argument: cognition doesn’t exclusively reside in the brain. It extends into the physical world when external objects integrate into cognitive processes.
They call it the „parity principle.“ If external processes are functionally equivalent to internal ones, we can’t draw arbitrary boundaries at skin and skull [30].
If that’s true, then worrying about „weakening“ cognition through tool use might be a category error. Like worrying that prosthetic limbs make you weak.
Some research supports this. The Education Endowment Foundation found that children who used calculators systematically throughout primary school had greater understanding and fluency with arithmetic [31].
Stanford noted generative AI can boost learning „for those who use it to engage in deep conversations and explanations“ [32].
Video games improve attention, reaction time, and spatial reasoning [33].
Risko and Gilbert’s 2016 review emphasizes that cognitive offloading isn’t inherently harmful. It benefits performance, particularly at higher memory loads [34]. People adaptively choose when to offload based on difficulty.
One robotics study found that cognitive offloading doesn’t prevent but rather promotes cognitive development [35]. Discovering reactive strategies actually facilitated more complex thinking.
Critics argue we’re overreacting. Vaughan Bell notes sensational claims about technology „make headlines across the world because they echo our recurrent fears“ despite lacking scientific support [36].
Christopher Ferguson points out: „If social media were causing an increase in teen suicides, we’d expect to see this pattern across countries with high technology adoption. But we don’t“ [37].
Maybe we’re just panicking. Again.
Every Generation Thinks the Kids Are Doomed
In Plato’s „Phaedrus,“ Socrates warned that writing would „create forgetfulness in the learners‘ souls, because they will not use their memories“ [3].
The irony is almost too perfect. Plato preserved these warnings through the very medium being criticized.
Writing enabled knowledge preservation across millennia. It’s literally the reason we know what Socrates thought.
Conrad Gessner might have been the first to raise alarms about information overload in the 16th century. He described printed material as „confusing and harmful to the mind“ [38].
Yet printing reversed the degradation of recorded thought. It enabled knowledge to compound and scale, catalyzing the Reformation, Renaissance, and Scientific Revolution [39].
The calculator debates of the 1970s echo current AI concerns almost word for word. Parents feared dependency and skill loss. Meta-analyses found that appropriate calculator use actually improves paper-and-pencil skills [40].
The scare was mostly a false alarm.
Amy Orben documents the „Sisyphean Cycle of Technology Panics“ [41]. Radio addiction in the 1940s. Comic book panics in the 1950s. Television. Video games. Social media. Same fears, different medium.
Douglas Adams nailed it: „Technology that existed when we were born seems normal, anything developed before we turn 35 is exciting, and whatever comes after that is treated with suspicion“ [42].
History suggests we overreact. A lot.
Conclusion
So what’s the verdict?
The evidence for concern is real. Cognitive offloading changes memory encoding [1]. Neural circuits atrophy without engagement [8]. Effortful learning is essential for durable knowledge [13]. The GPS research showing dose-dependent spatial memory decline is particularly damning [2].
But the counterarguments matter too. Historical technology panics routinely exceed actual harms [41]. Cognitive tools may extend rather than diminish our abilities when properly integrated [29]. And most studies can’t prove causation [17].
The synthesis? How we use technology matters more than whether we use it.
AI that removes cognitive effort undermines learning [16]. AI that promotes deep engagement can enhance it [32]. The distinction maps directly onto Bjork’s desirable difficulties framework: technology that removes struggle removes the mechanism of learning itself.
The generation effect, testing effect, and levels of processing research all confirm that struggle isn’t an obstacle to learning. It is learning [15].
In an AI age, the competitive edge may belong to those who cultivate deep expertise through effortful engagement rather than surface familiarity through convenient access.
The biological evidence is sobering. Neuroplasticity operates in both directions. The same mechanisms that strengthen the brain with use allow it to weaken with disuse [8].
Merzenich said it best: „It is just as easy to degrade the brain’s processing abilities as it is to strengthen or refine them“ [8].
The question isn’t whether we use technology. It’s whether we use it in ways that preserve and enhance cognitive capacity or allow it to atrophy.
The research suggests the answer depends not on the technology itself but on the consciousness and intentionality we bring to using it.
So maybe stop letting ChatGPT write all your emails. Your hippocampus will thank you.
References
[1] Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips. Science, 333(6043), 776-778. https://www.science.org/doi/10.1126/science.1207745
[2] Dahmani, L., & Bohbot, V. D. (2020). Habitual use of GPS negatively impacts spatial memory during self-guided navigation. Scientific Reports, 10, 6310. https://www.nature.com/articles/s41598-020-62877-0
[3] Plato. (c. 370 BCE). Phaedrus. Translated by Benjamin Jowett. https://www.perseus.tufts.edu/hopper/text?doc=Perseus:text:1999.01.0174:text=Phaedrus
[4] Wikipedia. Google effect. https://en.wikipedia.org/wiki/Google_effect
[5] Wegner, D. M. (1987). Transactive Memory: A Contemporary Analysis of the Group Mind. In B. Mullen & G. R. Goethals (Eds.), Theories of Group Behavior (pp. 185-208). Springer. https://en.wikipedia.org/wiki/Transactive_memory
[6] Google effects on memory: a meta-analytical review of the media effects of intensive Internet search behavior. (2024). BMC Psychology, 12, Article 30778. https://pmc.ncbi.nlm.nih.gov/articles/PMC10830778/
[7] Ward, A. F., Duke, K., Gneezy, A., & Bos, M. W. (2017). Brain Drain: The Mere Presence of One’s Own Smartphone Reduces Available Cognitive Capacity. Journal of the Association for Consumer Research, 2(2), 140-154. https://news.utexas.edu/2017/06/26/the-mere-presence-of-your-smartphone-reduces-brain-power/
[8] Merzenich, M. M. (2013). Soft-Wired: How the New Science of Brain Plasticity Can Change Your Life. Parnassus Publishing. https://www.kavliprize.org/the-brain-s-remarkable-capacity-for-change
[9] Kleim, J. A., & Jones, T. A. (2008). Principles of experience-dependent neural plasticity: implications for rehabilitation after brain damage. Journal of Speech, Language, and Hearing Research, 51(1), S225-S239.
[10] Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Good, C. D., Ashburner, J., Frackowiak, R. S., & Frith, C. D. (2000). Navigation-related structural change in the hippocampi of taxi drivers. Proceedings of the National Academy of Sciences, 97(8), 4398-4403. https://www.pnas.org/doi/10.1073/pnas.070039597
[11] Woollett, K., & Maguire, E. A. (2011). Acquiring „the Knowledge“ of London’s Layout Drives Structural Brain Changes. Current Biology, 21(24), 2109-2114. https://pmc.ncbi.nlm.nih.gov/articles/PMC3268356/
[12] Shors, T. J., Anderson, M. L., Curlik, D. M., & Nokia, M. S. (2012). Use it or lose it: How neurogenesis keeps the brain fit for learning. Behavioural Brain Research, 227(2), 450-458.
[13] Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher et al. (Eds.), Psychology and the real world: Essays illustrating fundamental contributions to society (pp. 56-64). Worth Publishers. https://bjorklab.psych.ucla.edu/research/
[14] Bjork, R. A., & Bjork, E. L. (2020). Desirable difficulties in theory and practice. Journal of Applied Research in Memory and Cognition, 9(4), 475-479. https://en.wikipedia.org/wiki/Desirable_difficulty
[15] Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592-604.
[16] Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2024). Generative AI Can Harm Learning. University of Pennsylvania Working Paper. https://knowledge.wharton.upenn.edu/article/without-guardrails-generative-ai-can-harm-education/
[17] Feeds, feelings, and focus: A systematic review and meta-analysis examining the cognitive and mental health correlates of short-form video use. (2025). Psychological Bulletin. https://www.psypost.org/large-meta-analysis-links-tiktok-and-instagram-reels-to-poorer-cognitive-and-mental-health/
[18] Richmond Functional Medicine. The Dopamine Effect: How Social Media Addiction Impacts Mental Health. https://www.richmondintegrative.com/blog/social-media-addiction
[19] Ophir, E., Nass, C., & Wagner, A. D. (2009). Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences, 106(37), 15583-15587.
[20] Newport, C. (2016). Deep Work: Rules for Focused Success in a Distracted World. Grand Central Publishing. https://calnewport.com/deep-work-rules-for-focused-success-in-a-distracted-world/
[21] McKinsey Global Institute. (2012). The social economy: Unlocking value and productivity through social technologies. https://knowledge.wharton.upenn.edu/article/deep-work-the-secret-to-achieving-peak-productivity/
[22] Leroy, S. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. Organizational Behavior and Human Decision Processes, 109(2), 168-181.
[23] Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow
[24] Fiske, S. T., & Taylor, S. E. (1991). Social Cognition (2nd ed.). McGraw-Hill.
[25] Research on calculator use and mathematical skills. ODU Digital Commons. https://digitalcommons.odu.edu/
[26] Research on spell-checker effects on learning. ResearchGate. https://www.researchgate.net/
[27] Microsoft Research & Carnegie Mellon. (2025). Study on AI confidence and critical thinking among knowledge workers.
[28] Bainbridge, L. (1983). Ironies of automation. Automatica, 19(6), 775-779.
[29] Clark, A., & Chalmers, D. (1998). The Extended Mind. Analysis, 58(1), 7-19. https://en.wikipedia.org/wiki/Extended_mind_thesis
[30] Clark, A., & Chalmers, D. (1998). The Extended Mind. Analysis, 58(1), 7-19. https://www.alice.id.tue.nl/references/clark-chalmers-1998.pdf
[31] Education Endowment Foundation. Calculator use in primary mathematics. https://educationendowmentfoundation.org.uk/
[32] Stanford Report on generative AI in education. (2024). https://hai.stanford.edu/
[33] Green, C. S., & Bavelier, D. (2003). Action video game modifies visual selective attention. Nature, 423(6939), 534-537.
[34] Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676-688.
[35] Robotics study on cognitive offloading and development. Journal of Cognitive Development.
[36] Bell, V. (2010). Don’t touch that dial! A history of media technology scares, from the printing press to Facebook. Slate. https://slate.com/technology/2010/02/a-history-of-media-technology-scares-from-the-printing-press-to-facebook.html
[37] Ferguson, C. J. (2017). Everything in moderation: Moderate use of screens unassociated with child behavior problems. Psychiatric Quarterly, 88(4), 797-805.
[38] Blair, A. (2003). Reading strategies for coping with information overload ca. 1550-1700. Journal of the History of Ideas, 64(1), 11-28.
[39] Eisenstein, E. L. (1980). The Printing Press as an Agent of Change. Cambridge University Press.
[40] Hembree, R., & Dessart, D. J. (1986). Effects of hand-held calculators in precollege mathematics education: A meta-analysis. Journal for Research in Mathematics Education, 17(2), 83-99.
[41] Orben, A. (2020). The Sisyphean Cycle of Technology Panics. Perspectives on Psychological Science, 15(5), 1143-1157.
[42] Adams, D. (1999). How to Stop Worrying and Learn to Love the Internet. The Sunday Times.